omics_update

This function allows to update UniProt IDs, add the sequence window, imputing NAs and calculating the Z-score.

To harmonize your raw omics data, you need to:

1. Read your input dataframe(s);
2. Format the input in the way described in the “Prepare your data” section;
3. If you provide a transcriptomics dataframe as input, choose the percent threshold of zeros to filter rows with >= threshold% of zeros;
4. If you provide a phosphoproteomics dataframe, choose the lenght of the sequence window. You can insert 7 or 15;
5. Choose the imputation method for proteomics and phosphoproteomics;
6. If you want to normalize with Z-score your data, insert TRUE for the zscore parameter, if not choose FALSE;
7. In the case you select TRUE for the Z-score calculation, you can also set the method (by “column” or by “row”) and the metric (“median” or “mean”).

This step for big dataframes requires many hours of working mainly due the imputation. For this tutorial we are using a smaller amount of patients to give you a general vision of

###Example of usage:

transcriptomics_data <- read_tsv("./PatientProfiler_raw_input/Brca_transc.tsv") 
proteomics_data <- read_tsv("./PatientProfiler_raw_input/Brca_prot.tsv") 
phosphoproteomics_data <- read_tsv("./PatientProfiler_raw_input/Brca_phospho.tsv")  

omics_update(
             df_tr = NULL,   
             df_pr = NULL,   
             df_ph = phosphoproteomics_data,  
             threshold = 80,
             sw_len = 7,  
             pep_col_name = "Peptide",
             imp_method = "norm",   
             zscore = TRUE,   
             zmethod = "column",   
             metric = "median",
             output_dir = 'PatientProfiler_processed_input'
             )

The result of this step is each selected dataframe manipulated according to the chosen parameters. If, for example, the phosphoproteomics update was made, it will have the information regarding phosphosites updated with the sequence_window column and Uniprot_ID added, it will be imputed, and it will have a distribution based on whether or not Z-score should be calculated.

After this step, tables are ready for the omics_preparation.

omics_preparation

For each uploaded omic type, a folder with a user-defined name will be created. Each folder will contain a file (data frame) for each patient, representing the needed omic input needed in SignalingProfiler-compliant format for the next steps.

Example of usage:

tr_updated <- read_tsv("./vignette/PatientProfiler_processed_input/Transcriptomics_updated.tsv")
pr_updated <- read_tsv("./vignette/PatientProfiler_processed_input/Proteomics_updated.tsv")
ph_updated <- read_tsv("./vignette/PatientProfiler_processed_input/Phosphoproteomics_updated.tsv")

omics_preparation(
              df_tr_updated = tr_updated,           
              df_pr_updated = pr_updated, 
              df_ph_updated = ph_updated,             
              transc_dir_name = "Transc_patients",   
              prot_dir_name = "Prot_patients",                  
              phospho_dir_name = "Phospho_patients"
              )

extract_protein_activity

This function is a wrapper to SignalingProfiler first step functions, wherein protein activity modulation is inferred from multi-omics data combining footprint-based techniques with PhosphoScore algorithm.

To extract protein activity for a single patient, you need to upload transcriptomics, proteomics, or phosphoproteomics data (or all of them) and set the parameters as described in the information part of the function (see the description by typing “?extract_protein_activity” in your console).

The parameters are managed as a list of default parameters that you can change as you wish, either inside or outside the function call.

(Note: The prot_df parameter in tf_params and kin_params will only be used if a prot_file is specified, since the function automatically loads proteomics data only if correct_proteomics is TRUE and if it finds the specified file)

Example of usage:

tf_params <- list(reg_minsize = 5, collectri = TRUE) 
kin_params <- list(reg_minsize = 5)

patient_id <- 'CPT000814'

extract_protein_activity(
                prot_file = paste0("Prot_patients/Prot_Patient_", patient_id, ".tsv"),       
                trans_file = paste0("Transc_patients/Transc_Patient_", patient_id, ".tsv"),   
                phospho_file = paste0("Phospho_patients/Phospho_Patient_", patient_id, ".tsv"), 
                tf_params,
                kin_params,
                output_dir = "Activities"
                )

Results output

For each patient an .xlsx file (Activity_patient_{patient_id}.tsv) will be saved inside the output_dir directory (default, ‘Activities’), with these columns:

• UNIPROT: protein Uniprot ID.

• gene_name: name of the protein.

• mf: molecular functions.

• final_score: activity score.

• method: how final_score is calculated.

extract_cohort_activity

To perform the protein activity inference step for all the patients of a cohort, this function loops on each omic file for each patient and infer protein activity using [extract_protein_activity].

(Note: The prot_df parameter in tf_params and kin_params will only be used if a prot_dir is specified, since the function automatically loads proteomics data only if correct_proteomics is TRUE and if it finds files in the specified folder)

tf_params <- list(reg_minsize = 5, collectri = TRUE)
kin_params <- list(exp_sign = TRUE)

extract_cohort_activity(
                        prot_dir = "Prot_patients/", 
                        trans_dir = "Transc_patients/",
                        phospho_dir = "Phospho_patients/", 
                        tf_params, 
                        kin_params,
                        output_dir = "Activities"
                        )

Results output

The cohort final results will be saved as a set of .tsv files (1 file - 1 patient=) inside the output_dir directory (default, ‘Activities’). Each patient-specific file (Activity_patient_{patient_id}.tsv) contains these columns:

• UNIPROT: protein Uniprot ID.

• gene_name: name of the protein.

• mf: molecular functions.

• final_score: activity score.

• method: how final_score is calculated.

create_network

This function sequentially calls five functions to:

1. Select a PKN derived from SIGNOR and PhosphoSitePlus (get_PKN);

2. Create a naive-network connecting mutations to inferred proteins (create_naive_network);

3. Optimize the naive network over protein activity (optimize_network_with_carnival);

4. Infer the activity of phenotypes from model proteins and connect them in a proteins-to-phenotypes network (infer_and_link_phenotypes);

5. Manipulate the proteins-to-phenotypes model for visualization and functional circuits creation (format_patient_networks).

Each step inside network creation requires a set of parameters, extensively described in the functions documentation (type ?initialize_net_default_params” in your console. The following function is called internally and automatically set up the default parameters, as identified in SignalingProfiler benchmarking. However, the user can customize the parameters for each step modifying PKN_options, naive_options, carnival_options, phenoscore_options, format_options parameters list.

output_dir <- 'Networks_output'
network_params <- initialize_net_default_params(output_dir)

Example of usage:

patient_id <- 'CPT000814'

# Read mutation file (starting nodes of the model)
mutations_df <- read_tsv('./PatientProfiler_raw_input/Brca_mutations.tsv', show_col_types = F)
sources <- mutations_df[mutations_df$Patient_ID == patient_id, ]

# Read activity file (intermediate nodes of the model)
activities <- read_tsv(paste0('./Activities/Activity_Patient_', patient_id, '.tsv'), show_col_types = F)

# Set the functional traits vector (final nodes of the model)
desired_phenotypes <- c('APOPTOSIS', 'PROLIFERATION')

# Read omics file
proteomics <- read_tsv(paste0('./Prot_patients/Prot_Patient_', patient_id, '.tsv'), , show_col_types = F)
transcriptomics <- read_tsv(paste0('./Transc_patients/Transc_Patient_', patient_id, '.tsv'), show_col_types = F)
phosphoproteomics <- read_tsv(paste0('./Phospho_patients/Phospho_Patient_', patient_id, '.tsv'), show_col_types = F)

# ProxPath proteins-to-phenotypes pre-processing
pheno_distances_table <- proxpath_preprocessing(proteomics = proteomics, 
                                                phosphoproteomics = phosphoproteomics)

create_network(patient_id = patient_id,
              sources = sources,
              activities = activities,
              transcriptomics = transcriptomics,
              proteomics = proteomics,
              phosphoproteomics = phosphoproteomics,
              desired_phenotypes = desired_phenotypes,
              pheno_distances_table = pheno_distances_table,
              output_dir = 'Networks_output',
              save_all_files = FALSE,
              PKN_options = list(direct = FALSE),
              naive_options = list(layers = 2, max_length = c(1,4)),
              carnival_options = list(carnival_type = 'vanilla_one_shot'),
              phenoscore_options = list(), # keep default params
              format_options = list(optimize_on_phenotypes = FALSE,
                                    circuits_params = list(k = -1),
                                    vis_cytoscape = FALSE))

create_cohort_network

This function loops on the folder of omics and activity data of a cohort of patients creating a network of each of the patients from mutations to desired phenotypes.

For the files generated for each patient, refer to create_network function documentation.

Example of usage:

create_cohort_networks(trans_dir = './Transc_patients/',
                      prot_dir = './Prot_patients/',
                      phospho_dir = './Phospho_patients/',
                      act_dir = './Activities/',
                      mut_file = './PatientProfiler_raw_input/Brca_mutations.tsv',
                      output_dir = './Networks_output/',
                      desired_phenotypes = c('APOPTOSIS', 'PROLIFERATION'),
                      pheno_distances_table = TRUE,
                      save_all_files = FALSE,
                      PKN_options = list(direct = FALSE),
                      naive_options = list(layers = 2, max_length = c(1,4)),
                      carnival_options = list(solver = 'cplex', 
                                              carnival_type = 'vanilla_one_shot'),
                                              carnival_params = list(timelimit = 3600),
                      format_options = list(optimize_on_phenotypes = FALSE,
                                    circuits_params = list(k = -1),
                                    vis_cytoscape = FALSE)
                      )

generate_communities

In this step, PatientProfiler extracts communities from the patient-specific mechanistic models generated in Step 3.

Importantly, this step requires a patient stratification file (patients_stratification.tsv), where each row represents a patient and the column indicates the subtype. You need to place this file in a designated directory (dir_path), which will be deleted after this step.

Example of usage:

generate_communities(dir_path = "./Communities/input_communities", 
                     network_dir = "./Networks_output",
                     output_dir = "./Communities/output_communities")

Results output

The output_dir directory contains a subdirectory for each detected community. Each community directory includes:

A file for edges. A file for nodes. A file listing the patients belonging to the community.

extract_signatures

Thanks to this function, we can extract signaling-driven transcriptomic signatures. To do this, you need the transcriptomics dataset obtained after Step 1 and the communities detected in Step 4.

Example of usage:

extract_signatures(base_path = "./Communities/output_communities/",          
                   transcriptomics_file = "./PatientProfiler_processed_input/Transcriptomics_updated.tsv",
                   output_dir = "./",
                   padj_thres = 0.01,
                   diff_thres = 0.5,
                   mean_exp_clus_thres = 0,
                   max_val = 50)
  
  

Results output

The Signatures directory contains a dataframe for each signature with the genes list.